Interface MultimodalPrompt

All Known Implementing Classes:
MiniMaxMultimodal, QwenVlProcessor.ProcessedMultimodal
public interface MultimodalPrompt
A prompt whose media (images, videos) was preprocessed by a model-specific processor. The engine only needs the expanded token ids and whether vision tensors are present; the ModelExecutor that produced the prompt interprets the rest.

  • Method Summary

    Modifier and Type
    Method
    Description
    boolean
    hasVision()
    Returns whether the prompt carries vision tensors.
    int[]
    inputIds()
    Returns the prompt token ids with every media placeholder expanded to its final length.

  • Method Details

    • inputIds

      int[] inputIds()
      Returns the prompt token ids with every media placeholder expanded to its final length.
      Returns:
      token ids.

    • hasVision

      boolean hasVision()
      Returns whether the prompt carries vision tensors.
      Returns:
      true when media must be encoded during prefill.